Model Marketplace

doubao-seedance-2-0-mini-260615

Seedance 2.0 mini is a new-generation cost-effective video generation model for broader video generation needs. While maintaining competitive quality, it brings video generation capability to lower-threshold, higher-frequency, and more scalable application scenarios.

glm-5.2

GLM-5.2 is a new-generation open-source flagship model designed for Long Horizon Tasks by Zhipu AI, supporting 1M lossless ultra-long context. With outstanding programming and engineering capabilities, it can autonomously complete the full pipeline from task decomposition, architecture design, front-end/back-end development, integration testing to multi-platform deployment, suitable for complex engineering tasks, long-horizon interaction, code generation, enterprise applications and other scenarios.

happyhorse-1.1-i2v

HappyHorse-1.0-T2V supports text-to-video generation, featuring highly realistic dynamic rendering. It accurately comprehends text semantics to produce high-quality videos that are fluid, natural, and rich in detail.

happyhorse-1.1-r2v

HappyHorse-1.0-T2V supports text-to-video generation, featuring highly realistic dynamic rendering. It accurately comprehends text semantics to produce high-quality videos that are fluid, natural, and rich in detail.

happyhorse-1.1-t2v

HappyHorse-1.0-T2V supports text-to-video generation, featuring highly realistic dynamic rendering. It accurately comprehends text semantics to produce high-quality videos that are fluid, natural, and rich in detail.

qwen-image-2.0-pro

Qwen-Image-2.0 series full-power version model, achieving fusion of image generation and image editing; with more professional text rendering supporting 1K token instructions, more delicate realistic texture for realistic scenes, stronger semantic compliance. The full-power version has the strongest text rendering capability and realistic texture in the 2.0 series.

qwen-image-2.0

Qwen-Image-2.0 series accelerated version model, achieving fusion of image generation and image editing; with more professional text rendering supporting 1K token instructions, more delicate realistic texture for realistic scenes, stronger semantic compliance. The accelerated version effectively achieves the best balance between model quality and performance.

qwen3.7-plus

Qwen3.7 series high cost-effectiveness Plus model, comprehensively upgrading visual-language capabilities based on strong text capabilities while maintaining complete agent capabilities in coding, tool usage, and productivity workflows. Its core feature is multimodal interaction hybrid agent capability, able to perceive real-world scenes, read screens and operate GUI, generate code based on visual references, and navigate mobile applications end-to-end.

MiniMax-M2.5

MiniMax-M2.5 is a flagship open-source large model by MiniMax. Through large-scale reinforcement learning training in hundreds of thousands of real complex environments, M2.5 reaches or refreshes industry SOTA in programming, tool calling and search, office and other productivity scenarios.

Hy3 preview

Hunyuan Hy3 preview is designed for Agent workloads, adopting 295B/21B activated MoE architecture. Within the same model, it provides three modes: no_think (ultra-fast response), think_low (fast thinking), think_high (deep reasoning), adapting to different latency and depth needs from high-frequency interaction to complex engineering tasks. Approaching current best levels on code benchmarks like SWE-bench Verified, 256K context supports cross-file code refactoring and long document analysis. Suitable for developers who need reliable task completion while being sensitive to inference costs.

qwen3.7-max

Qwen3.7 series largest and most capable Max model, currently offering pure text model capabilities for experience. Qwen3.7 is a new-generation flagship model for the agent era, with core advantages in the breadth and depth of agent capabilities: excelling at programming, office and productivity, and long-cycle autonomous execution tasks.

happyhorse-1.0-t2v

HappyHorse-1.0-T2V supports text-to-video generation, featuring highly realistic dynamic rendering. It accurately comprehends text semantics to produce high-quality videos that are fluid, natural, and rich in detail.

happyhorse-1.0-r2v

HappyHorse-1.0-I2V enables image-to-video generation, featuring highly realistic dynamic rendering. It accurately comprehends both text and image semantics to produce high-quality videos that are fluid, natural, and rich in detail.

happyhorse-1.0-i2v

HappyHorse-1.0-I2V enables image-to-video generation, featuring highly realistic dynamic rendering. It accurately comprehends both text and image semantics to produce high-quality videos that are fluid, natural, and rich in detail.

glm-5.1

GLM-5.1 is a model designed for Long Horizon Tasks by Zhipu AI, with total parameters 744B, supporting 200K ultra-long context, maximum output 128K tokens. With powerful logical reasoning, long text understanding and code generation capability, balancing performance and inference efficiency; excellent performance in multi-task benchmarks, suitable for intelligent interaction, enterprise applications, development assistance and other scenarios.

deepseek-v4-pro

Flagship MoE large model, total parameters 1.6T, activated 49B, natively supporting million-level ultra-long context. With massive high-quality training data, featuring top-tier math logic, complex reasoning, professional code and long text deep analysis capability, adapting to high-difficulty scenarios like advanced research, complex office, and deep intelligent agents.

deepseek-v4-flash

High-efficiency lightweight MoE model, total parameters 284B, activated 13B, natively supporting million-token ultra-long context. Fast reasoning speed, low latency, low calling cost, balanced comprehensive capability, focusing on high-concurrency lightweight tasks, suitable for daily dialogue, content creation, basic RAG, batch copywriting and other essential scenarios.

qwen-flash-character

Qwen series multilingual role-playing model. This model is a dynamically updated version with advance notification of updates. Suitable for anthropomorphic role-playing, with optimized persona instruction following, topic advancement, and empathetic listening capabilities, supporting deep restoration of personalized characters.

kimi-k2.5

kimi-k2.5 is the most versatile model released by Moonshot AI to date, with native multimodal architecture design, supporting both visual and text input, thinking and non-thinking modes, dialogue and Agent tasks.

wan2.7-r2v

Wanxiang 2.7 Reference-to-Video, more stable character, prop, and scene reference, supporting up to 5 image/video mixed references, audio timbre reference, with upgraded basic capabilities for stronger performance.